Re: IDS Next Version - Idea status?

Urich Ann wrote: > ? What is the status of some of the suggestions that were made at the > 2004 IBM Tech Conference (Birds of Feather) such as: > > 1. Create a catalog table to hold "explain" information. > Whenever an "explain" command is executed, the access > plan results would be sent to this table. > This would allow history of the optimizer's access > path... > allowing a compare of the rows to see if > the code (access path) has changed. I've responded separately in detail. To my considerabl surprise, I couldn't find this FR in PTS (though, had I been asked, I'd've said I'd created it a few years back), so I created one for it. And the number is at the office -- IIRC, it is 173270. No immediate plans to deal with it, even though it is a good idea. > 2. Create a Visual-Explain similar tool. > Basically a tool that would give a graphical > presentation > and/or sql-tuned recommendations to the SQL being > explained. > (Use the current IBM visual explain tool's code and pump > in > Informix's explain output/access path info.) Again perfectly reasonable - but not on the schedule or in PTS until I created FR 173271 (give or take). > 3. Ability to access all catalog tables by "tabname" instead of > a "tabid join". > Creating "views" of the Catalog tables to allow "catalog > table" access via "tabname". > ... (Instead of having to join on "tabid" for every > single catalog query.) > Example: > A view of the sysindexes table (which is called.... > "ibm.sysindexes".) > This view defined (behind the scenes) joining > sysindexes and systables by tabid... > We (DBA's) could then write: > > Select idxname from ibm.sysindexes > where tabname = 'customer_table';> > Instead of.... > > Select idxname from sysindexes ix, systables tb > where tabname = 'customer_table' > and ix.tabid = tb.tabid; Not sure about this one. A single view isn't a big problem (though MODE ANSI database with multiple tables with the same name but different owners are somewhat problematic). However, if you combine two such views, you are messing with self-joins of (or, at least, multiple joins to) systables, which is messy. I didn't create an FR for this. If we were going to do something about it (heads up, Ann - new info) then it would most likely be along the lines of the SQL/Schema stuff in ISO/IEC 9075-11:2003. There's an archaic preview of that in the xpg4.sql file in $INFORMIXDIR/etc. It likely runs into the same issues with multiple uses of systables where only one is needed if you join separate views. -- Jonathan Leffler #include <disclaimer.h> Email: jleffler@earthlink.net, jleffler@us.ibm.com Guardian of DBD::Informix v2005.01 -- http://dbi.perl.org/

"Jonathan Leffler" <jleffler@earthlink.net> wrote in message > I couldn't find this FR in PTS ... WTF?

Captain Pedantic wrote: > "Jonathan Leffler" <jleffler@earthlink.net> wrote in message > > >>I couldn't find this FR in PTS ... > > > WTF? > ...seems like IBM finally got to Jonathan. He as been assimilated. FeatureRequest and PheTure System ??? ;-) Cheers Serge -- Serge Rielau DB2 SQL Compiler Development IBM Toronto Lab

Captain Pedantic wrote: > "Jonathan Leffler" <jleffler@earthlink.net> wrote in message > > >>I couldn't find this FR in PTS ... > > WTF? No - it doesn't mean *that*! :-) Apologies - FR is feature request, and PTS is Problem Tracking System. A quick Google shows that I've more often than not been remiss in expanding on the acronym (a dozen entries since 2000-01-01 mention PTS, and only one self-evidently expands the acronym). FR I've expanded more often - and the Google search is less clear-cut since it appears as the Internet country code TLD for France, and as part of 'rm -fr', etc. -- Jonathan Leffler #include <disclaimer.h> Email: jleffler@earthlink.net, jleffler@us.ibm.com Guardian of DBD::Informix v2005.01 -- http://dbi.perl.org/